Gemma 4 Hackathon Banner
Official 2026 Hackathon

Gemma 4 Hackathon | 2026

Launched by GDG China, the Gemma 4 Hackathon challenges the developer community to build real AI applications with cutting-edge capabilities. Beyond technology itself, the focus is solving real-world problems.

Status

Three-city Semifinals

Finalist Slots

15-30 Teams

Award Support

Up to $200,000 resource support

Offline Finals

2026 Google I/O Connect China

View Timeline

Event Overview

This edition focuses on Gemma 4 across four core tracks so teams can choose a clear direction and go deep: Track A - AI Agent (multi-step planning, tool use, and native function calling for autonomous assistants such as medical pre-screening, legal document workflows, industrial inspection, and education coaching); Track B - Multimodal (integrated vision, audio, and language experiences for scenarios like cultural heritage digitization, accessibility tools, and mixed-format document Q&A); Track C - Edge AI (fully offline deployment on phones, Raspberry Pi, or embedded hardware with real device demos, such as rural healthcare assistants, privacy-preserving local OCR, and low-power agriculture monitoring); and Track D - AI for Social Good (solutions for accessibility, climate, public health, education equity, and disaster response). Whether you prefer cloud orchestration, multimodal innovation, edge deployment, or social impact, you can find your main battleground here.

Through rigorous technical architecture evaluation and social impact assessment, the competition will select 15-30 high-potential projects with strong innovation and execution value. Top teams will be invited to present at the Google I/O Connect final demo stage, compete for official recognition, and gain deep technical support and global enablement from the Google ecosystem.

Track Overview

AI Agent Track

Track A: AI Agent

Build autonomous AI agents with Gemma 4 capabilities in multi-step planning, tool usage, and native function calling.

Examples: medical pre-screening assistant, legal document agent, industrial inspection agent, and education tutoring bot.

Multimodal Track

Track B: Multimodal

Deeply integrate vision, audio, and language to deliver truly multimodal experiences beyond simple model chaining.

Examples: cultural heritage digitization, tools for visually impaired users, and mixed-format document Q&A.

Edge AI Track

Track C: Edge AI

Hardcore challenge: fully offline deployment of E2B/E4B-class models on phones, Raspberry Pi, or embedded hardware (real device demo required).

Examples: rural healthcare assistant, privacy-preserving local OCR, and low-power crop disease monitoring.

AI for Social Good Track

Track D: AI for Social Good

Use Gemma 4 to address social challenges such as accessibility, climate, public health, education equity, and disaster response.

Examples: disability assistance tools, climate monitoring, mental health self-help, and humanitarian translation.

Finalists

Congratulations to the teams advancing to the final round, grouped by track (15 teams).

Finalist information submission deadline: July 14, 2026, 23:59

Information submission is now closed.

Track A: AI Agent

4 teams
Finalist

Team Name: SuperMap

Project Name: GIS Data Agent

Finalist

Team Name: 妙妙爱学习

Project Name: 时间规划小助手

Finalist

Team Name: 得法团队

Project Name: MedSchema Agent：基于 Gemma 4 的专病指标清洗与质控智能体

Finalist

Team Name: 晓话AI

Project Name: MonkeyAI

Track B: Multimodal

2 teams
Finalist

Team Name: OOtira Team

Project Name: OOtira — Privacy-First AI Closet Copilot powered by Gemma 4 On-Device Vision

Finalist

Team Name: 躺平别动队

Project Name: 乐学

Track C: Edge AI

4 teams
Finalist

Team Name: enactflow

Project Name: attractor

Finalist

Team Name: PandaaX

Project Name: 守灵犀 · EdgeMind

Finalist

Team Name: The Weeknd

Project Name: 智能盆栽

Finalist

Team Name: 妙手回春

Project Name: FirstAid Copilot

Track D: AI for Social Good

5 teams
Finalist

Team Name: AI for HER

Project Name: AI for HER

Finalist

Team Name: AI向善

Project Name: Gemma4多模态深度赋能：跨境AI短视频刻板印象智能检测与包容引导应用

Finalist

Team Name: Elva

Project Name: Elva 老白 - 老人AI声导伴侣

Finalist

Team Name: KnifShield

Project Name: SafeVoice Agent: Elderly Anti-Fraud Voice AI Agent

Finalist

Team Name: TerraVineVox

Project Name: VoxShield — AI 声纹守护者

Timeline

Plan your 2026 timeline now

  1. Kickoff & Registration (Apr 18)

    Official launch, registration opens, partner onboarding

  2. Workshop Series (Apr 18 - May 18)

    Five online workshops covering model foundations, multimodal, agent building, and edge deployment

  3. Registration Deadline (May 18)

    Team lock-in (1-5 members), cross-city teams welcome

  4. Hackathon Build Phase (May 18 - Jun 8)

    Open development period

  5. Mentor Support (May 18 - Jun 8)

    Experts and GDEs provide online office hours

  6. Final Submission (Jun 8, 23:59)

    Submit code repository + demo video (within 5 minutes) + technical report + online demo URL

  7. Three-City Semi-Finals (Jun 13)

    In-person semi-final roadshows at Google offices in Beijing, Shanghai, and Shenzhen

  8. Finalist Announcement (Jul 15)

    Finalist list published

  9. National Grand Final

    Live final demo at 2026 Google I/O Connect China, competing for top honors!

Awards & Judging Criteria

Award Categories

  • Grand Champion (1)

    20,000 Cloud Credit (shared across all participating teams), plus GDE nomination opportunity and Google for Startups collaboration opportunities

  • Track Winner Awards (3)

    Dedicated awards per track

  • Best Mobile App Award (1) and Best Edge Deployment Award (1)

  • Excellence Awards (11)

Judging Dimensions

  1. Real-world impact (30%):Addresses real needs, with clear target users and scalability.
  2. Technical excellence (25%):Code quality, architecture design, and depth of Gemma 4 usage.
  3. Completeness (20%):Runnable demo, complete core functionality, and robust edge-case handling.
  4. Innovation (15%):Novel approach, creative model usage, and unique problem framing.
  5. Presentation quality (10%):Clear technical report, professional video, and solid code documentation.

Rules

Core Model

The core model must be Gemma 4 (any size/spec is allowed).

Compliance

All training data sources must be disclosed and comply with data privacy laws.

Resources & Support

To help teams move faster, we prepared:

  • Expert Mentors:GDE mentors provide end-to-end guidance.
  • Q&A Channels:Office Hours + WeChat support group.

Learning Resources & Tools

Why join the Gemma Hackathon?

  • Guidance from Google teams and direct exchange on cutting-edge practices
  • Top teams can showcase at 2026 Google I/O Connect and gain major visibility
  • Get direct feedback from expert judges and investors to accelerate commercialization
  • Meet cross-functional collaborators and build under real hackathon intensity
  • Use AI to create real social value

Who should join this hackathon?

  • Developers, UI/UX designers, PMs, and operators who want to build startup projects
  • Startup teams building in the Google ecosystem
  • Individual developers and teams interested in Google AI

FAQ

Q1: Can I register as an individual?

Yes. You may participate solo or form a team later.

Q2: How many teams pass initial screening?

We plan to select 15-30 teams for the next stage.

Q3: Where is the final held?

In August at 2026 Google I/O Connect, with live demos and awards.

Q4: Is participation guaranteed after registration?

All registrations are reviewed. Teams that best match themes and rules will be prioritized.

Q5: I registered but did not receive a confirmation, and my team is not listed on the website. What should I do?

Please contact @指导老师-陈老师 in the community chat group.